Abstract
Multi-agent software platforms increasingly assign each autonomous AI agent ("persona") a first-class email identity so the agent can correspond like a human teammate. Doing this at scale exposes four under-addressed engineering problems: (1) provisioning APIs typically demand an explicit per-tenant domain setup step before any account exists, coupling agent creation to tenant onboarding; (2) an address must be synthesized deterministically and collision-safely from an agent's human-readable identity; (3) modern JMAP mail servers expose a provisioning-time account identifier distinct from the session-time account identifier used on the JMAP wire — conflating them silently breaks mail; and (4) the app-passwords minted during provisioning live in process memory and are lost on restart, silently disabling all agent mail until manual repair.
This publication discloses a unified method resolving all four: lazy idempotent domain creation (created once on first agent, no-op thereafter); slug-based local-part synthesis with tenant-scoped collision detection; dual-identifier resolution that persists the provisioning identifier while resolving the session identifier on-demand from a JMAP session call and caching it; and credential rehydration that reloads every persisted credential from an encrypted store into the in-process cache at boot, so mail works immediately post-restart without re-provisioning. The combination delivers working mailboxes within seconds of agent creation and high post-restart availability with zero manual steps. We disclose this publicly and defensively to keep the technique freely practiceable and to bar later patenting by others.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Assuncao, gustavo matthew, "Persona Mailbox Auto-Provisioning with Credential Rehydration", Technical Disclosure Commons, ()
https://www.tdcommons.org/dpubs_series/10580